Overview
- CEO Dario Amodei said the company cannot accept new contract language he described as allowing safeguards to be bypassed, even as talks continue up to the 5:01 p.m. ET Friday cutoff.
- Defense Secretary Pete Hegseth set the deadline and warned of canceling Anthropic’s roughly $200 million contract, designating the firm a supply‑chain risk, or invoking the Defense Production Act.
- Pentagon officials say they seek use of Claude for all lawful purposes, deny plans for mass surveillance or autonomous weapons, and argue legality is the end user’s responsibility.
- The Pentagon has contacted major contractors to assess reliance on Anthropic ahead of a possible supply‑chain risk label, while xAI’s Grok has agreed to classified use and other labs are in advanced talks.
- Claude has been the only AI model cleared on some classified networks, and experts cited in coverage question whether the Defense Production Act could be used to force changes to Anthropic’s safeguards.